More Cache - What good is it?

All about them.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

More Cache - What good is it?

Post by ces » Wed Apr 14, 2010 8:08 am

What attributes of performance does extra cache help?

Here are Newegg's Dual Core CPU prices:
E3300 Wolfdale 2.50GHz 1MB L2 Cache $51.99
E5400 Wolfdale 2.70GHz 2MB L2 Cache $69.99
E6500 Wolfdale 2.93GHz 2MB L2 Cache $79.99
E6600 Wolfdale 3.06GHz 2MB L2 Cache $97.99
E7400 Wolfdale 2.80GHz 3MB L2 Cache $124.99
E8400 Wolfdale 3.00GHz 6MB L2 Cache $167.99


1. Low Cost - What uses would cause me to not choose the e3300?

2. More Cache - What uses would cause me to choose the E8400 for the extra cache? How much extra $ is extra 5MB pf cache worth?
How about the E7400? How much extra $ is extra 2MB pf cache worth? What good is it really?
How about the E5400? How much extra $ is extra 1MB pf cache worth? What good is it really?

3. More GHz - What uses would cause me to choose the E6500 for the extra GHz? How much extra $ is the extra 0.43GHz worth?
Last edited by ces on Wed Apr 14, 2010 6:11 pm, edited 1 time in total.

Jay_S
*Lifetime Patron*
Posts: 715
Joined: Fri Feb 10, 2006 2:50 pm
Location: Milwaukee, WI

Post by Jay_S » Wed Apr 14, 2010 12:55 pm

Google is your friend...

Legion Hardware from 2007: Intel Conroe Cache Performance.
Tom's Hardware from 2007: Does Cache Size Really Boost Performance?
Anandtech from 2008: Nehalem Part 3: The Cache Debate, LGA-1156 and the 32nm Future

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Wed Apr 14, 2010 2:45 pm

Jay_S wrote:Google is your friend...

Legion Hardware from 2007: Intel Conroe Cache Performance.
Tom's Hardware from 2007: Does Cache Size Really Boost Performance?
Anandtech from 2008: Nehalem Part 3: The Cache Debate, LGA-1156 and the 32nm Future
There is a lot of data there, but I think I need a human being who understands that data and convert it into a practical assessment.

The benchmarks are synthetic, about gaming, or about encoding. None of which applies to me. What I use a lot are browsers, Microsoft office, windows explorer, bloated scanning software, and a PDF reader.

Are people like me that much rarer then people using obtuse encoding or decoding software? There has to be a lot of MS Office users out there - that is where MS gets the largest portion of its revenue.

Games and non-name software that less then 5% of all users use.

I am hoping someone who understands what those benchmarks are actually measuring can translate the obtuse benchmarks into a more accessible reality.

It is beyond me to be able to do that.

judge56988
Friend of SPCR
Posts: 455
Joined: Sun Jul 27, 2008 10:39 am
Location: England

Post by judge56988 » Wed Apr 14, 2010 10:31 pm

ces wrote:
There is a lot of data there, but I think I need a human being who understands that data and convert it into a practical assessment.

The benchmarks are synthetic, about gaming, or about encoding. None of which applies to me. What I use a lot are browsers, Microsoft office, windows explorer, bloated scanning software, and a PDF reader.
If that's all you use a computer for, I don't think you need to worry too much about about how much cache you've got or how fast your memory is.
Computers were coping quite happily with those kind of tasks 5 years ago and more, when DDR3 and dual core processors were not available.
Something like a Athlon II X4 630 or Core i3 530 and a couple of gigs of RAM would be fine!
Spend your money on good quality quiet components and chill a little.

As to waiting for the next great development before you buy something; as has been said by many people many times, you'd never buy anything and probably still be using log tables and a slide rule.:lol:
There is always something better around the corner and it'll no doubt have even more power that you don't need.
It's all about hype and marketing - constant upgrading means more money for the companies that make these components.

nutball
*Lifetime Patron*
Posts: 1304
Joined: Thu Apr 10, 2003 7:16 am
Location: en.gb.uk

Post by nutball » Wed Apr 14, 2010 11:30 pm

ces wrote:I am hoping someone who understands what those benchmarks are actually measuring can translate the obtuse benchmarks into a more accessible reality.

It is beyond me to be able to do that.
OK, here's the precis:

- stop worrying about it

Memory speed and cache size/speeds will have no practical impact on the performance of a PC for the uses you are describing.

For the uses you're describing it's not even clear that there'd be a real noticeable performance difference between any of the CPUs you listed in your OP, clock speed, cache or anything else. Only difference is cost.

Vicotnik
*Lifetime Patron*
Posts: 1831
Joined: Thu Feb 13, 2003 6:53 am
Location: Sweden

Re: More Cache - What good is it?

Post by Vicotnik » Thu Apr 15, 2010 3:08 am

ces wrote: E3300 Wolfdale 2.50GHz 1MB L2 Cache $51.99

1. Low Cost - What uses would cause me to not choose the e3300?
Any reason why you're only listing Intel CPUs? AMD is king of the low end and based on what you're going to use the system for I'd say that an unlocked Sempron 140 or an inexpensive Athlon II X2, X3 or X4 would do nicely.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Re: More Cache - What good is it?

Post by ces » Thu Apr 15, 2010 3:16 am

Vicotnik wrote:
ces wrote: E3300 Wolfdale 2.50GHz 1MB L2 Cache $51.99

1. Low Cost - What uses would cause me to not choose the e3300?
Any reason why you're only listing Intel CPUs? AMD is king of the low end and based on what you're going to use the system for I'd say that an unlocked Sempron 140 or an inexpensive Athlon II X2, X3 or X4 would do nicely.
Two reasons.
1. I am not familiar with the AMD line (but maybe I should better educate myself about it)
2. I already own the Zotac mini-ITX 775 board. So my socket decision is already made.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Thu Apr 15, 2010 3:19 am

judge56988 wrote:
ces wrote:
There is a lot of data there, but I think I need a human being who understands that data and convert it into a practical assessment.

The benchmarks are synthetic, about gaming, or about encoding. None of which applies to me. What I use a lot are browsers, Microsoft office, windows explorer, bloated scanning software, and a PDF reader.
If that's all you use a computer for, I don't think you need to worry too much about about how much cache you've got or how fast your memory is.
Computers were coping quite happily with those kind of tasks 5 years ago and more, when DDR3 and dual core processors were not available.
Something like a Athlon II X4 630 or Core i3 530 and a couple of gigs of RAM would be fine!
Spend your money on good quality quiet components and chill a little.

As to waiting for the next great development before you buy something; as has been said by many people many times, you'd never buy anything and probably still be using log tables and a slide rule.:lol:
There is always something better around the corner and it'll no doubt have even more power that you don't need.
It's all about hype and marketing - constant upgrading means more money for the companies that make these components.
I already own the Zotac mini-ITX 775 board. I bought a Q9550 for it. But I want to power it with the winmate PSU system and while I might be able to do that, I don't want to cut it so close.

So if I am going to use a smaller 775 socket chip, I am trying to figure out what it should be.

judge56988
Friend of SPCR
Posts: 455
Joined: Sun Jul 27, 2008 10:39 am
Location: England

Post by judge56988 » Thu Apr 15, 2010 4:02 am

ces wrote: So if I am going to use a smaller 775 socket chip, I am trying to figure out what it should be.
Well, if it were me and I was positive that I didn't want to do any serious gaming, video editing or 3D CAD work; I'd go for the E5400 and stick the Quad core on e-bay.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Thu Apr 15, 2010 4:03 am

The Limitations of Today's CPU Benchmarks

There is a problem with all the benchmarks out there. They don't appear to translate well to what an average user uses. They seem to be built around the curiosity of: (a) gamers, (b) high definition enthusiasts, and (c) hardware enthusiasts interested in synthetic benchmarks to chase.

My Typical Everyday Use

I don't know that I am typical, but I am more typical than the above. Here is what I use. Typically with a fair number of windows open at the same time:
Browsers (Chrome and Foxfire and sometimes MS internet explorer)
Windows Explorer (I use it a lot and it sometimes has pauses while it waits for my HDD to be doing something - a black edition 1T WD)
Word processing (MS Office & Open Office - both can sometimes be slow in launching a file through Windows explorer)
Scanning software (bloated third party, bloated Brother, and extremely bloated HP software)
Anti-Virus (AVG and sometimes Avira - I have switched portions of them off because they can drain system resources)
X1 desktop search engine (when it is indexing files it loads the system up similarly to the Anti-virus software)
iDrive backup software (when it is backing up the system it loads the system similarly to the Anti-virus software)
PDF (Foxit PDF reader, Adobe PDF reader and CoolPDF printer)
farstone cloning/backup software

My Experience

For some time I have run the above applications on an Intel e2200 system with a raptor boot drive.

When I upgraded to an Intel E8200 with a velociraptor boot drive andblack series WD data drive, if it was faster, it wasn't faster by much.

I recently got a Clarkdale 650 system with a vertex SSD boot drive and a black series WD data drive, I haven't really move to it yet, but this one seems to feel faster, but not by as much as I would have liked.

What I am Doing

I am building a new computer. I have a Zotac mini-ITX board. I have a Lian Li Q07 case. I was going to put a Q9550 CPU in it. But I want to power it with ElectroDacus's winmate PSU system. I am not fully comfortable using it to power the Q9550.

So I am looking for a less power hungry substitute 775 cpu. The published benchmarks do not seem to be helpful.

Can You Help Me?

Can you help me translate the data in the available benchmarks to help me evaluate what CPU to use on this Zotac 775 system.

After using the Clarkdale 650, I wonder if the Q9550 is going to be a better performer. The 650 with multi-threading and additional two virtual cores, is probably the equivalent of a three core chip. It's cores run faster than the Q9550 by a lot, not even counting Turbo.

So I don't think this 775 system is going to be a speed upgrade to the Clarkdale system. But I don't want a sluggish system either.

So what CPU do I select? What is the tradeoffo in Cache and in GHz for my uses?
Last edited by ces on Thu Apr 15, 2010 6:05 am, edited 1 time in total.

judge56988
Friend of SPCR
Posts: 455
Joined: Sun Jul 27, 2008 10:39 am
Location: England

Post by judge56988 » Thu Apr 15, 2010 4:24 am

ces wrote: Can You Help Me?
I can tell you a few things that I've been told/read, (sorry if it's a grandma and egg sucking thing, but hey.)

How many services do you have running in Task Manager? - Turn off anything you don't need.
Defragment your hard drive on a regular basis and get rid of crap you don't need - temp files, browser cache, unused programmes and so on.
Do your back-ups and virus scanning when you are not using the computer. (Based on the amount of posts you do, that's probably at night time :wink: )
Pause the indexing service when you are working.
Fix the size of your swap file, this keeps it in one place on your HDD and cuts down on fragmentation.
Keep the registry clean.

I'm far from an expert on computers, most of these tips I picked up were to avoid dropping frames doing video capture; now I actually use a dual boot system with one system stripped to the barest minimum needed. This is my video "workstation" and is much more responsive than the normal everyday system.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Thu Apr 15, 2010 6:07 am

judge56988 wrote:
ces wrote: Can You Help Me?
How many services do you have running in Task Manager? - Turn off anything you don't need.
Defragment your hard drive on a regular basis and get rid of crap you don't need - temp files, browser cache, unused programmes and so on.
Do your back-ups and virus scanning when you are not using the computer. (Based on the amount of posts you do, that's probably at night time :wink: )
Pause the indexing service when you are working.
Fix the size of your swap file, this keeps it in one place on your HDD and cuts down on fragmentation.
Keep the registry clean.
I have done some of that. It probably has helped. But that begs the question. Which 775 cpu do I choose and why. And how do I make sense out of the cpu benchmark data that is out there in order to assist in making this decision?

judge56988
Friend of SPCR
Posts: 455
Joined: Sun Jul 27, 2008 10:39 am
Location: England

Post by judge56988 » Thu Apr 15, 2010 6:22 am

nutball wrote:
For the uses you're describing it's not even clear that there'd be a real noticeable performance difference between any of the CPUs you listed in your OP, clock speed, cache or anything else. Only difference is cost.
'nuff said.

ilovejedd
Posts: 676
Joined: Tue Oct 21, 2008 4:14 pm
Location: in the depths of hell

Post by ilovejedd » Thu Apr 15, 2010 6:28 am

ces wrote:I have done some of that. It probably has helped. But that begs the question. Which 775 cpu do I choose and why. And how do I make sense out of the cpu benchmark data that is out there in order to assist in making this decision?
Quite frankly, any of them will work, even the lowly Celeron E3300. There's a reason why most benchmarks are geared towards enthusiasts. It's because processors nowadays are more than capable of meeting regular users' demands. The most noticeable upgrade you can likely make is a fast SSD.

For heavy multi-tasking, a quad-core might help, however, I still don't know what's defined as heavy. I'm not familiar with a lot of the applications you mentioned but multi-window, multi-tab Firefox (often more than 200 tabs open) with Microsoft Security Center running in the background works fine on my PC (Zotac GF9300-G-E/Celeron E3300 2.5GHz/4GB DDR2 800).

Jay_S
*Lifetime Patron*
Posts: 715
Joined: Fri Feb 10, 2006 2:50 pm
Location: Milwaukee, WI

Post by Jay_S » Thu Apr 15, 2010 6:43 am

ces wrote:There is a problem with all the benchmarks out there. They don't appear to translate well to what an average user uses. They seem to be built around the curiosity of: (a) gamers, (b) high definition enthusiasts, and (c) hardware enthusiasts interested in synthetic benchmarks to chase.
Our hypothetical average user has no idea what cache is. The average user confuses memory and storage, thinks Windows comes with Office, and thinks POP is a beverage. They are not the intended audience for tech sites.

I agree - synthetics do not translate into real-world performance. Most tech sites use canned benchmarks because it's a convenient way to compare hardware. But others don't - for example, [H]ardOCP eschews canned benchmarks and instead actually plays games to test hardware. They report on the gaming experience. They're a PC gaming enthusiast site, and their tests translate perfectly into what their intended audience does.
ces wrote:Here is what I use. Typically with a fair number of windows open at the same time:
Browsers (Chrome and Foxfire and sometimes MS internet explorer)
Windows Explorer (I use it a lot and it sometimes has pauses while it waits for my HDD to be doing something - a black edition 1T WD)
Word processing (MS Office & Open Office - both can sometimes be slow in launching a file through Windows explorer)
Scanning software (bloated third party, bloated Brother, and extremely bloated HP software)
Anti-Virus (AVG and sometimes Avira - I have switched portions of them off because they can drain system resources)
X1 desktop search engine (when it is indexing files it loads the system up similarly to the Anti-virus software)
iDrive backup software (when it is backing up the system it loads the system similarly to the Anti-virus software)
PDF (Foxit PDF reader, Adobe PDF reader and CoolPDF printer)
farstone cloning/backup software
I'd say you typify the average office PC user. Most of your slowdowns sound like I/O issues, and will not be helped by larger CPU cache. The benchmarks related to cache were picked because they are CPU-intensive. Ie, the disks are waiting for the CPU to complete whatever it's doing. I don't think any of your aps behave this way.

I would get rid of AVG. AVG used to be the light-weight AV of choice, but has become bloated and slow over the past few years. I dropped AVG as soon as they started with the advert pop-ups (8? 8.5?). I use MS Security Essentials, which has been terrific to date. It's lightweight, fast, and well-reviewed.

Lots of windows, browsers and browser tabs will eat RAM. If you're using a 64-bit OS, don't be afraid to go a little nuts on RAM. If you're not overclocking, don't worry about memory clock speed. But pay attention to latencies. Get low-latency JDEC-compliant RAM, and lots of it.

The rest of your issues sound like
1) poor software coding
2) I/O limitations of your mechanical disks
ces wrote:I am building a new computer. I have a Zotac mini-ITX board. I have a Lian Li Q07 case. I was going to put a Q9550 CPU in it. But I want to power it with ElectroDacus's winmate PSU system. I am not fully comfortable using it to power the Q9550.

So I am looking for a less power hungry substitute 775 cpu. The published benchmarks do not seem to be helpful.
The three articles I linked to were a response to your question: "More cache, what good is it?", and is no help for the power question. The least power-hungy LGA775 CPU right now is probably one of the Wolfdale Celerons (xbit labs) precisely because they have less cache. I have an E3300 + G41 board in my HTPC, and it consumes no more power than the previous AMD 4050e (45W TDP) + 740G/SB700 system it replaced.
ces wrote:So I don't think this 775 system is going to be a speed upgrade to the Clarkdale system. But I don't want a sluggish system either.

So what CPU do I select? What is the tradeoffo in Cache and in GHz for my uses?
I don't think the LGA775 system will be faster than the Clarkdale system either. Most of your applications are not heavily threaded, so the Q9550's extra cores don't benefit you. Your aps look I/O bound. Faced with that I'd spend on a fast SSD and - to a lesser extent - high CPU clock speed for good single-threaded performance.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Thu Apr 15, 2010 7:35 am

Jay_S wrote:I would get rid of AVG. AVG used to be the light-weight AV of choice, but has become bloated and slow over the past few years. I dropped AVG as soon as they started with the advert pop-ups (8? 8.5?). I use MS Security Essentials, which has been terrific to date. It's lightweight, fast, and well-reviewed.
MS Security Essentials???? That is against my religion.

So you think single core speed should be the driving consideration. That would make one of these a good choice:
E6500 Wolfdale 2.93GHz 2MB L2 Cache $79.99
E6600 Wolfdale 3.06GHz 2MB L2 Cache $97.99

Though neither of these would be a bad choice:
E3300 Wolfdale 2.50GHz 1MB L2 Cache $51.99
E5400 Wolfdale 2.70GHz 2MB L2 Cache $69.99

And I should focus on an SSD boot drive. I don't think I am ready to trust my data to an SSD though.

psyopper
Friend of SPCR
Posts: 336
Joined: Fri Mar 28, 2008 1:15 pm
Location: Portland, Oregon

Post by psyopper » Thu Apr 15, 2010 8:00 am

First, it seems your bottlenecks are related to a crappy OS. If all you are doing is surfing the web, scanning for viruses, indexing your drive and using OpenOffice then I might suggest a change of operating systems.

Simply moving to Linux removes the overhead of scanning for viruses and indexing your drive. OpenOffice works just dandy on Linux, as do Firefox, Chrome, and even IE under Wine.

Edit - I also forgot to mention you were overlooking the obvious answer to your quad core quandry... the Q9550S - a 65w version of the same processor: http://www.newegg.com/Product/Product.a ... 6819115210

Also you could consider undervolting your current 9550 and getting the same results: http://www.silentpcreview.com/intel-q9550s

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Thu Apr 15, 2010 8:08 am

psyopper wrote:First, it seems your bottlenecks are related to a crappy OS. If all you are doing is surfing the web, scanning for viruses, indexing your drive and using OpenOffice then I might suggest a change of operating systems.

Simply moving to Linux removes the overhead of scanning for viruses and indxing your drive. OpenOffice works just dandy on Linux, as do Firefox, Chrome, and even IE under Wine.
I know that linux is a better operating system. It is overlaod for the human equation (me) though.

And openoffice is good for a number of uses. But it has some weaknesses that require I be able to use MS Word. Trying to operate stuff under wine is just too steep of an investment of personal time.

So as much as I dislike MS, I am stuck with them.

ilovejedd
Posts: 676
Joined: Tue Oct 21, 2008 4:14 pm
Location: in the depths of hell

Post by ilovejedd » Thu Apr 15, 2010 8:34 am

ces wrote:MS Security Essentials???? That is against my religion.
You're using Windows. I don't really get the aversion to Microsoft Security Essentials when it's free, fast, lightweight and seems to work better than other options.

Jay_S
*Lifetime Patron*
Posts: 715
Joined: Fri Feb 10, 2006 2:50 pm
Location: Milwaukee, WI

Post by Jay_S » Thu Apr 15, 2010 8:43 am

ces wrote:MS Security Essentials???? That is against my religion.
May I ask why? Just because MS is the company we all love to hate? :)
ces wrote:So you think single core speed should be the driving consideration. That would make one of these a good choice:
E6500 Wolfdale 2.93GHz 2MB L2 Cache $79.99
E6600 Wolfdale 3.06GHz 2MB L2 Cache $97.99

Though neither of these would be a bad choice:
E3300 Wolfdale 2.50GHz 1MB L2 Cache $51.99
E5400 Wolfdale 2.70GHz 2MB L2 Cache $69.99
Between those, there's a few differences beyond clock speed, cache size and price:
http://ark.intel.com/Compare.aspx?ids=4 ... 2771,42805,
ces wrote:And I should focus on an SSD boot drive. I don't think I am ready to trust my data to an SSD though.
Data on mechanical storage will take just as long to open (Word files, PDF's, etc). So you're I/O bound again. What about this: keep current or often-used data on the SSD, with backups on mechanical storage.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Thu Apr 15, 2010 9:36 am

Jay_S wrote:
ces wrote:MS Security Essentials???? That is against my religion.
May I ask why? Just because MS is the company we all love to hate? :)
Yeah sort of. Everyone has their own reasons, I have mine.

Each person's particular reasons may differ, but they all rhyme.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Thu Apr 15, 2010 9:40 am

Jay_S wrote: Data on mechanical storage will take just as long to open (Word files, PDF's, etc). So you're I/O bound again. What about this: keep current or often-used data on the SSD, with backups on mechanical storage.
Yeah I will do something like that. I do like to keep my data on another drive. I have a silverstone device that lets you use an SSD as cache for a hard drive. That is an option. I can also create a small 1st partition on a large 1T or 2T drive that will be fast. I have some old raptor drives.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Thu Apr 15, 2010 9:45 am

Theoretical Question

All things equal, theoretically, what should provide a more responsive system for the type of use I am talking about:
1. a single core chip operating at 4Ghz
2. a dual core chip operating at 2GHz
3. a quad core chip operating at 1Ghz


4. Does the performance ranking between these chips change as cache is added or removed from such chips (assume the same amount of cache per core).

ilovejedd
Posts: 676
Joined: Tue Oct 21, 2008 4:14 pm
Location: in the depths of hell

Post by ilovejedd » Thu Apr 15, 2010 10:43 am

ces wrote:Theoretical Question

All things equal, theoretically, what should provide a more responsive system for the type of use I am talking about:

1. a single core chip operating at 4Ghz
2. a dual core chip operating at 2GHz
3. a quad core chip operating at 1Ghz


4. Does the performance ranking between these chips change as cache is added or removed from such chips (assume the same amount of cache per core).
I think a dual-core at 2GHz should make for a happy medium. A lot of apps are sufficiently multi-threaded to take advantage of dual-core and 2GHz is a decent clock speed. The 4GHz single-core should theoretically work fastest (no overhead) but having two cores helps with multi-tasking.

The benefit of more cache and higher FSB is dependent on the specific application. So far, the ones you've mentioned are all I/O-bound (disk-bound) making SSD's the best upgrade you can make. You'll likely see very little difference between the earlier processors mentioned.

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Thu Apr 15, 2010 10:49 am

ilovejedd wrote:I think a dual-core at 2GHz should make for a happy medium. A lot of apps are sufficiently multi-threaded to take advantage of dual-core and 2GHz is a decent clock speed. The 4GHz single-core should theoretically work fastest (no overhead) but having two cores helps with multi-tasking.

The benefit of more cache and higher FSB is dependent on the specific application. So far, the ones you've mentioned are all I/O-bound (disk-bound) making SSD's the best upgrade you can make. You'll likely see very little difference between the earlier processors mentioned.
Wouldn't a single core at 4GHz be fastest in an IO bound situation - where it is competing with one of 4 cores operating at 1Ghz?

Jay_S
*Lifetime Patron*
Posts: 715
Joined: Fri Feb 10, 2006 2:50 pm
Location: Milwaukee, WI

Post by Jay_S » Thu Apr 15, 2010 11:08 am

ces wrote:Wouldn't a single core at 4GHz be fastest in an IO bound situation - where it is competing with one of 4 cores operating at 1Ghz?
By "I/O bound", we mean your disks - not the CPU. Mechanical storage is slow, and even a 1 GHz quad will be waiting for storage. Hence the SSD recommendation. Specifically, Intel SSD's shine at random I/O, which happens to typify average PC usage.

As you may well know, clock speed is only 1 part of the whole package. As AMD showed the world back in the Netburst era, clock speed is a poor indicator of overall CPU performance. "Work per tick" is equally important to raw frequency. Like how the Core i7 920 immediately obsoleted the Extreme edition Core 2 quad at lower frequency and 1/3 the price in certain applications.

Even if all your applications are single-threaded, the fact that any/every PC has multiple applications/processes running all the time means they'll benefit from extra cores.

Vicotnik
*Lifetime Patron*
Posts: 1831
Joined: Thu Feb 13, 2003 6:53 am
Location: Sweden

Post by Vicotnik » Thu Apr 15, 2010 11:21 am

ces wrote:Theoretical Question

All things equal, theoretically, what should provide a more responsive system for the type of use I am talking about:
1. a single core chip operating at 4Ghz
2. a dual core chip operating at 2GHz
3. a quad core chip operating at 1Ghz


4. Does the performance ranking between these chips change as cache is added or removed from such chips (assume the same amount of cache per core).
The way I see it higher frequency gives more raw power. More cores gives more responsiveness. However a single core at 4X GHz will almost always beat a dual core at 2X GHz and a quad core at X GHz in both those areas, so the choices you're presenting are a bit odd. A more realistic comparison would be looking at real processors and their cost.

Many heavy duty stuff you're likely to do will be run in a single thread, or will be the kind of stuff that can be parallelized but not that well. How long the process takes then greatly depends on the speed of that single core used.

However when a heavy task is running it's nice to have lots of space CPU power so that the system doesn't feel sluggish.

So it depends on how you're using your computer really. For me a quad core at a moderate frequency is perfect since I seldom do heavy stuff that I have to wait on due to the CPU - most of the time the I/O system is the bottleneck (network bandwidth, HDD speed, etc) but those times that I do run something heavy I want the system to work great at the same time.
If I were a gamer I think a faster dual core would suit my needs better.

And an SSD is great for responsiveness. Get one! :)

nutball
*Lifetime Patron*
Posts: 1304
Joined: Thu Apr 10, 2003 7:16 am
Location: en.gb.uk

Post by nutball » Thu Apr 15, 2010 11:46 am

ces wrote:Theoretical Question

All things equal, theoretically, what should provide a more responsive system for the type of use I am talking about:
1. a single core chip operating at 4Ghz
2. a dual core chip operating at 2GHz
3. a quad core chip operating at 1Ghz
Things aren't equal, because those aren't the products being offered.

Seriously ces, you're over-analysing the problem, and worse still you're over-analysing the wrong problem.

psyopper
Friend of SPCR
Posts: 336
Joined: Fri Mar 28, 2008 1:15 pm
Location: Portland, Oregon

Post by psyopper » Thu Apr 15, 2010 11:54 am

In reality there's a thousand web sites with a thousand benchmarks of each of those processors. Why not do a bit of research and take a look at the comparisons that are equivelant to the type of work you are going to be doing and see where they stack.

Something tells me that the E8400 will always score the highest (it's half of a Q9550), but it's the ratio performance:cost that you are interested in. For each of the benchmarks you can ask yourself "Is that worth $110 to me?"

ces
Posts: 3395
Joined: Thu Feb 04, 2010 6:06 pm
Location: US

Post by ces » Thu Apr 15, 2010 1:23 pm

You may be able to read and interpret the benchmarks. I have tried and have not been able to.

Sort of like a medical patient looking at his xray or a legal client reading a 70 page complaint. Merely reading it isn't of any use. You have to be able to read, understand and interpret it.

Post Reply